06 June 2024 : Review article
The Potential of Artificial Intelligence in Prosthodontics: A Comprehensive Review
Ibrahim Saleh Aljulayfi
1ABC*, Ali Hamoud Almatrafi2ABD, Ramzi O. Althubaitiy1ACG, Fahad Alnafisah3BCF, Khalid Alshehri3BCF, Bandar Alzahrani3BCF, Khalid Gufran 4DEG
DOI: 10.12659/MSM.944310
Med Sci Monit 2024; 30:e944310
Table 5 Selected studies for prosthodontic treatment planning and prosthesis designing.
| Authors | AI Model | Application | Datasets | Results | Conclusions |
|---|---|---|---|---|---|
| Chau R et al (2022) []30 | GAN | Design of the single molar dental prosthesis by AI | 250Patients | – | AI can automate the design of single-tooth dental prostheses and identify the features of the remaining dentition |
| Chen Q et al (2016) []33 | Knowledge-based clinical support system with CBR and cosine algorithm | Designing removable partial dentures. Designs of RPDs for 104 randomly selected patients were compared with those selected by professionals | 104 patients | AUC-ROC (96%), mean average precision (61%) | Knowledge-based clinical support system is efficient in RPD design. Because the domain knowledge in dentistry has the same logic and representation format, other methods can also be applied to other fields in dentistry |
| Chen Y et al (2022) []34 | Knowledge-based AI (CEREC, BI) | Designing lithium disilicate dental crowns. Comparing the occlusal morphology and fracture behavior of lithium disilicate ceramic dental crowns on 12 human participants’ premolar #45 designed by a knowledge-based AI (CEREC, biogeneric individual function, BI). Designed crowns in 3 groups for comparison (AI, experienced technician, and trained dental students) | 12 teeth | Occlusal profile discrepancy (0.3677±0.0388) | Designing lithium disilicate dental crowns, CAD design with humans may be better than knowledge-based AI |
| Matin et al (2017) []35 | Expert system for the simulation model design and manufacturing | Casting metal substructure for designing the metal-ceramic crown | Common data model approach, blackboard architecture, rule-based reasoning, and iterative redesign method. | The mean value of arithmetic roughness on cast substructure (1935 to 2776 μm) | The time required for manufacturing the metal substructure using the Expert system is shorter than the time required for manufacturing without the expert system |
| Mine et al (2020) []37 | Artificial Neural Network (ANN) | Variegation in maxillofacial prosthesis fabrication. Finding the appropriate amount of pigment by contrasting two machine learning algorithms: the random forest algorithms and ANN-based deep learning | A spectropho-meter was used to evaluate the CIE 1976 L* a* b* color space information on 52 silicone elastomeric specimens of different hues using the input dataset | Color difference (3.45±0.87) | When compared to the random forest algorithm, the deep ANN approach yielded better results in terms of the ΔE00 value |
| Otani T et al (2015) []36 | Robotics | Automatic (robotic) tooth preparation for porcelain laminate veneers | 20 maxillary central incisors | Accuracy (=0.15), precision (=0.30), standard deviation (=0.034) | The control process was able to prepare the tooth model with greater precision than the experimental procedure, but both methods were able to prepare the tooth model similarly precisely. At the finish line, the experimental group exhibited higher levels of precision and accuracy |
| Zhang et al (2019) []38 | CNN | Extract of tooth preparation margin line | 380 models of dental preparations | Accuracy (97.43%), specificity (97.59%), sensitivity (97.32%) | The study revealed that CNN automatically accomplishes the extraction of the tooth preparation margin line accurately |
| Tian et al (2021) []31 | GAN | Design of inlay restorations | 750 dental prostheses | – | GAN can accurately segment an internal surface. Experiments on a real-world database show that the GAN model outperforms the traditional methods, which can restore the groove shape consistently, with the residual tooth surface |
| Tian et al (2022) []32 | GAN | Reconstruction of the occlusal surface | 1000 patients | Error rate (0.114), standard deviation (0.195) | GAN outperforms state-of-the-art approaches in occlusal surface reconstruction. Importantly, the developed occlusal surface has sufficient anatomical shape of actual teeth and great clinical application value |
| Chau et al (2023) []42 | GAN | Examine the accuracy of a GAN in constructing biomimetic dental prostheses for single molars | 169 casts | True reconstruction (60%) | Accuracy of biomimetic GAN-designed dental prostheses could be further enhanced |
| Choi et al (2023) []43 | DL | Extract of tooth preparation marginal finish line | 182 casts | True reconstruction (100%) | Deep learning and computer-aided design approaches enable the robust and accurate extraction of finish lines |
| Ding et al (2023) []44 | 3D-DCGAN | Use of machine learning algorithms to design a dental crown | 600 casts | Root mean square value (0.3611) | 3D-DCGAN could be utilized to design personalized dental crowns with high accuracy that can mimic both the morphology and biomechanics of natural teeth |
| Farook et al (2023) []45 | 3D-CNN | Development and evaluation of three-dimensional convolutional neural network (3D-CNN) to produce partial dental crowns | 30 specimens | Accuracy (60%), sensitivity (100%), precision (83%) | 3D-CNN can design and generate partial dental crowns in CAD for restorative dentistry |
| Liu (2024) []46 | AI | Evaluate the clinical applicability of AI to design dental restorations | 15 specimens | Surface truth (68.4 μm) | AI can assist in the production of dental restorations, thereby enhancing both production efficiency and accuracy |
| AI – artificial intelligence; GAN – generative adversarial network; CBR – case-based research; RPD – removable partial denture; CEREC – chair-side economical restoration of esthetic ceramic; BI – business intelligence; CNN – convolutional neural network; 3D-DCGAN – 3D-deep convolutional generative adversarial network; PSNR – peak signal-to-noise ratio. | |||||